This paper examines the problem of data placement in Bubba, a highly-parallel system for data-intensive applications being developed at MCC. “Highly-parallel” implies that load balancing is a critical performance issue. “Data-intensive” means data is so large that operations should be executed where the data resides. As a result, data placement becomes a critical performance issue. In general, determining the optimal placement of data across processing nodes for performance is a difficult problem. We describe our heuristic approach to solving the data placement problem in Bubba. We then present experimental results using a specific workload to provide insight into the problem. Several researchers have argued the benefits of declustering (i e, spreading each base relation over many nodes). We show that as declustering is increased, load balancing continues to improve. However, for transactions involving complex joins, further declustering reduces throughput because of communications, startup and termination overhead. We argue that data placement, especially declustering, in a highly-parallel system must be considered early in the design, so that mechanisms can be included for supporting variable declustering, for minimizing the most significant overheads associated with large-scale declustering, and for gathering the required statistics.

Abstracf-Since 1984, the goal of the Bubba project at MCC has been to design a scalable, high-performance and highly available database system that will provide significant costlperformance advantages over conventional mainframes in the 1990's. The design process has been an iterative one, cycling through design, modeling, and prototyping in progressive detail. The current Bubba prototype runs on a commercial 40-node multicomputer and includes a parallelizing compiler, distributed transaction management, object management, and a customized version of UNIX. This paper describes the current prototype and discusses of the major design decisions that went into its construction. The lessons learned from this prototype and its predecessors are presented.Index Terms-Complex object management, database operating system, database programming language, database system performance, database system prototype, dataflow execution, parallel database system.

This paper examines the problem of data placement in Bubba, a highly-parallel system for data-intensive applications being developed at MCC. "Highly-parallel" implies that load balancing is a critical performance issue. "Data-intensive" means data is so large that operationsshould be executed where the data resides. As a result, data placement becomes a critical performance issue.In general, determining the optimal placement of data across processing nodes for performance is a difficult problem. We describe our heuristic approach to solving the data placement problem in Bubba. We then present experimental results using a specific workload to provide insight into the problem. Several researchers have argued the benefits of declustering (i.e., spreading each base relation over many nodes). We show that as declustering is increased, load balancing continues to improve. However, for transactions involving complex joins, further declustering reduces throughput because of communications, startup and termination overhead.We argue that data placement, especially declustering, in a highly-parallel system must be considered early in the design, so that mechanisms can be included for supporting variable declustering, for minimizing the most significant overheads associated with large-scale declustering, and for gathering the required statistics.

An accurate method to measure the four point probe (4PP) sheet resistance (R s ) of USJ Source-Drain structures is described. The new method utilizes Elastic Material probes (EMprobe) to form non-penetrating contacts to the silicon surface. The probe design is kinematic and the force is controlled to ensure elastic deformation of the probe material. The probe material is selected so that large direct tunneling currents can flow through the native oxide thereby forming a low impedance contact. Sheet resistance measurements on USJ implanted P+/N structures with SIMS junction depths as shallow as 15 nm have been measured. The sheet resistance values obtained with the new EM-probe 4PP method were found to be consistent with expectations. In this paper, the method will be demonstrated on a variety of implanted USJ structures.

The effects of geometry, contact misalignment and sheet resistivity on the extracted values of specific contact resistance have been simulated for both circular and square contacts. These resuits have been used to detail the errors involved in extracting contact resistivity from Kelvin structures with rectangular and circular contacts.